To reduce traffic accidents it is important to reveal the mechanisms behind them. Therefore we collected data from events that nearly lead to an accident. We found out that, among all these incidents, near rear-end collisions were the most common type and that they often occur when following a preceding car. Using an incident database we found special circumstances leading to these rear-end incidents. Analyzing the traffic circumstance and the behavior of the drivers in these situations a model for the probable reaction of the drivers was developed. Furthermore the influence of adjacent cars was quantified by performing experiments with a driving simulator based on this model.